The Research On Stereo Matching Technology In Binocular Stereo Vision System

Computer vision is an interdisciplinary subject involving computer science,pattern recognition,image processing and artificial intelligence,and binocular stereoscopic vision is one of the focuses of computer vision.Stereo matching is a key step in binocular stereoscopic vision,which plays an important role in binocular stereoscopic vision.This paper mainly studies the stereoscopic matching theory in binocular vision.This dissertation first introducing the working principle of binocular stereo vision system and the basic theory of stereo matching for the different elements of matching,the region-based and feature-based stereo matching methods are improved respectively —— a stereo matching algorithm based on region and stereo matching algorithm based on edge feature are proposed.The main achievements of this paper are as follows:Firstly,in the region-based stereo matching algorithm,each reference point of the image is matched by block-matching and the parameters are adjusted adaptively according to the number of suspicious matching points.Thus,the preliminary suspicious matching point set is obtained.Secondly,the 8 neighborhood gray features to filter the suspicious matching points are adopted to get the most possible set of matching points.Finally,the final matching point is tested by the consistency constraint of stereo matching to get the best matching point.In the stereo matching algorithm based on edge feature,first performing edge detection on the matched color image,the dissertation divides the image into edge part,edge adjacent part and non-edge part.Firstly,six characteristic attributes are proposed for the non-edge part and the neighboring edge part of the image.The initial matching image is obtained by comprehensive evaluation of the six characteristic attributes.Secondly,the edge part of the image is matched according to the matching result of the non-edge part and the neighboring edge part of the image.Finally,by correcting the parallax,the final matching image is obtained.The experimental results show that the algorithm proposed in this paper can achieve the stereo matching with high accuracy.It can not only get a very close parallax image,but also effectively achieve the three-dimensional scene reconstruction.